Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for generating higher quality visual data from lower quality visual data using fidelity transfer, the method comprising: receiving input visual data representing at least one image, the input visual data including lower quality visual data; selecting fidelity visual data from a database of fidelity visual data based on similarity with the lower quality visual data, the fidelity visual data having a resolution higher than the lower quality visual data; extracting features from the lower quality visual data; and estimating, by a hierarchical algorithm, higher quality visual data from the lower quality visual data and the fidelity visual data, the higher quality visual data having a resolution higher than the lower quality visual data, wherein the estimating includes transposing the fidelity visual data onto content of the extracted features of the lower quality visual data.
This invention relates to enhancing the quality of visual data, such as images or video, by improving resolution and fidelity. The problem addressed is the challenge of upgrading lower-quality visual data (e.g., low-resolution images) to higher-quality outputs while preserving or enhancing visual details. The method involves receiving input visual data, which may be degraded in quality, and selecting reference fidelity visual data from a database based on similarity to the input. The fidelity visual data has a higher resolution than the input. Features are extracted from the lower-quality input, and a hierarchical algorithm processes both the input and the fidelity data to generate higher-quality visual output. The algorithm transposes the fidelity data onto the content of the extracted features, effectively transferring high-resolution details from the reference data to the input. This approach leverages pre-existing high-quality visual data to improve the resolution and visual fidelity of lower-quality inputs, addressing limitations in traditional upscaling techniques. The hierarchical algorithm ensures that the enhancement process is structured and adaptive, optimizing the transfer of fidelity characteristics while maintaining coherence with the original content.
2. The method of claim 1 , wherein the similarity is determined using a cosine similarity function.
A system and method for analyzing data similarity involves comparing data elements to determine their similarity. The method calculates a similarity score between two data elements by applying a cosine similarity function. Cosine similarity measures the angle between two vectors in a multi-dimensional space, providing a value between -1 and 1, where 1 indicates identical vectors, 0 indicates orthogonal vectors, and -1 indicates opposite vectors. This approach is particularly useful in applications such as document retrieval, recommendation systems, and clustering, where understanding the relationship between data points is critical. The cosine similarity function is chosen for its efficiency in high-dimensional spaces and its ability to handle sparse data, making it suitable for large-scale data analysis. The method may involve preprocessing the data, such as normalizing or scaling, to ensure accurate similarity measurements. The resulting similarity score can then be used to group similar data elements, identify outliers, or rank data based on relevance. This technique is widely applied in natural language processing, image recognition, and machine learning to enhance data-driven decision-making.
3. The method of claim 1 , wherein the fidelity visual data includes a plurality of features, wherein the plurality of features are computed by a neural network trained on object recognition.
The invention relates to a method for processing visual data to enhance object recognition. The method involves generating fidelity visual data that includes multiple features, which are computed using a neural network specifically trained for object recognition tasks. The neural network processes input visual data to extract and identify relevant features that improve the accuracy and reliability of object recognition. These features may include spatial, texture, or shape-based characteristics that help distinguish objects within the visual data. The method leverages the neural network's learned representations to enhance the fidelity of the visual data, making it more suitable for applications requiring precise object detection and classification. This approach improves the robustness of object recognition systems by utilizing advanced machine learning techniques to analyze and interpret visual information. The method is particularly useful in fields such as autonomous vehicles, surveillance, and augmented reality, where accurate and efficient object recognition is critical.
4. The method of claim 3 , wherein the plurality of features of the fidelity visual data include one or more content features and one or more style features.
This invention relates to a method for processing visual data to enhance its fidelity, addressing the challenge of maintaining both content and stylistic integrity in digital images or video. The method involves analyzing visual data to extract multiple features, including content features that represent the core subject matter and style features that define aesthetic or artistic characteristics. These features are then used to improve the overall quality and realism of the visual data, ensuring that both the content and style are preserved or enhanced. The method may involve techniques such as feature extraction, data augmentation, or machine learning-based processing to refine the visual data. By separating and independently processing content and style features, the method allows for more precise adjustments, such as enhancing details while maintaining the original artistic style. This approach is particularly useful in applications like image restoration, style transfer, or high-fidelity rendering, where preserving both the subject matter and artistic intent is critical. The method can be applied to various types of visual data, including photographs, digital art, or video frames, to produce outputs that are both visually accurate and stylistically consistent.
5. The method of claim 1 , wherein the database includes a plurality of feature vectors, wherein the selecting the fidelity visual data includes: computing a feature vector for the lower quality visual data, the feature vector configured to approximate a feature vector of the higher quality visual data; and comparing the feature vector to the plurality of feature vectors of the database to obtain the fidelity visual data.
This invention relates to a method for selecting visual data with improved fidelity by leveraging a database of feature vectors. The problem addressed is the challenge of efficiently identifying and retrieving high-fidelity visual data from a dataset, particularly when working with lower-quality input data. The method involves computing a feature vector for the lower-quality visual data, where this feature vector is designed to approximate the feature vector of the higher-quality visual data. This computed feature vector is then compared against a plurality of feature vectors stored in a database. The comparison process identifies the most relevant or matching fidelity visual data from the database, ensuring that the selected visual data closely resembles the higher-quality reference. The database of feature vectors serves as a reference library, enabling accurate and efficient retrieval of high-fidelity visual data based on the computed approximation. This approach enhances the accuracy and reliability of visual data selection, particularly in applications where lower-quality input data is common, such as image processing, computer vision, and multimedia retrieval systems.
6. The method of claim 5 , wherein the comparing includes maximizing a cosine similarity between the feature vector and one or more of the plurality of feature vectors included in the database.
This invention relates to a method for comparing feature vectors in a database, particularly in applications such as machine learning, data retrieval, or pattern recognition. The method addresses the challenge of efficiently and accurately matching a feature vector against a collection of stored feature vectors to identify the most relevant or similar entries. The method involves comparing a feature vector against multiple feature vectors stored in a database. The comparison process specifically maximizes the cosine similarity between the input feature vector and one or more of the stored feature vectors. Cosine similarity is a measure of similarity between two vectors, calculated as the cosine of the angle between them, which is particularly useful for high-dimensional data where Euclidean distance may be less meaningful. By maximizing this similarity, the method ensures that the most relevant or closely related vectors are identified, improving the accuracy of retrieval or classification tasks. The method may be applied in various domains, such as recommendation systems, image recognition, natural language processing, or any application requiring efficient similarity-based searches. The use of cosine similarity ensures robustness against variations in vector magnitudes, focusing instead on the directional similarity between vectors. This approach enhances computational efficiency and accuracy in matching tasks, making it suitable for large-scale databases and real-time applications.
7. The method of claim 1 , wherein the higher quality visual data includes additional visual data that is not included in the lower quality visual data.
This invention relates to a method for enhancing visual data quality in a system where lower quality visual data is initially captured or transmitted, and higher quality visual data is subsequently obtained. The method addresses the problem of limited bandwidth or processing constraints that prevent the immediate capture or transmission of high-quality visual data, such as in video streaming, surveillance, or remote sensing applications. The method involves capturing or receiving lower quality visual data, which may have reduced resolution, frame rate, or color depth compared to the higher quality visual data. The higher quality visual data includes additional visual information not present in the lower quality version, such as finer details, higher resolution, or additional color channels. The method processes the lower quality visual data to identify areas where additional visual information is needed, then integrates the higher quality visual data into these areas to produce an enhanced output. The system may use techniques such as super-resolution, interpolation, or machine learning-based enhancement to improve the lower quality visual data before integrating the higher quality data. The method ensures that the final output retains the additional visual information from the higher quality data while minimizing artifacts or inconsistencies. This approach is particularly useful in applications where real-time processing is required, and higher quality data cannot be transmitted or stored immediately.
8. The method of claim 1 , wherein the estimating includes estimating the higher quality visual data based on a cost function, wherein the cost function includes a regularization with a probabilistic model fitted to the fidelity visual data.
This invention relates to methods for enhancing visual data quality, particularly in scenarios where lower-quality visual data is available alongside higher-fidelity reference data. The problem addressed is the challenge of accurately estimating higher-quality visual data from lower-quality inputs, such as in image or video processing, where noise, compression artifacts, or resolution limitations degrade the visual fidelity. The method involves estimating higher-quality visual data by using a cost function that incorporates regularization. The regularization is based on a probabilistic model that has been fitted to the available fidelity visual data. This probabilistic model helps constrain the estimation process, ensuring that the reconstructed higher-quality data adheres to known statistical properties of the reference data. The cost function balances the trade-off between faithfulness to the input data and adherence to the learned probabilistic model, improving the accuracy and realism of the enhanced output. The approach is particularly useful in applications like medical imaging, surveillance, or multimedia processing, where improving visual quality while preserving critical details is essential. By leveraging probabilistic modeling, the method avoids overfitting to noise or artifacts, resulting in more reliable and visually coherent reconstructions. The technique can be applied to various types of visual data, including images, videos, or 3D reconstructions, where higher-quality outputs are desired from lower-quality inputs.
9. The method of claim 1 , wherein the features of the lower quality visual data are extracted using the hierarchical algorithm.
This invention relates to a method for processing visual data, particularly for extracting features from lower quality visual data using a hierarchical algorithm. The method addresses the challenge of accurately analyzing or interpreting visual data that is degraded, noisy, or otherwise of poor quality, which can occur in applications such as surveillance, medical imaging, or low-light photography. The hierarchical algorithm processes the visual data in multiple stages, progressively refining the extracted features to improve accuracy and robustness. The algorithm may include steps such as initial coarse feature detection, followed by iterative refinement to capture finer details. This approach allows the system to handle varying levels of data quality while maintaining reliable feature extraction. The method may be applied in systems where real-time processing is required, such as autonomous vehicles or augmented reality devices, where low-quality visual inputs are common. The hierarchical structure ensures that even partially degraded data can yield useful features, enhancing the overall performance of visual analysis tasks.
10. The method of claim 1 , wherein the hierarchical algorithm includes a neural network, the neural network includes a convolution neural network.
A method for processing data using a hierarchical algorithm that incorporates a neural network, specifically a convolutional neural network (CNN), to improve computational efficiency and accuracy. The hierarchical algorithm organizes data into multiple levels of abstraction, allowing the CNN to analyze features at different scales and resolutions. The CNN processes input data through convolutional layers, which apply filters to extract spatial hierarchies of features, followed by pooling layers to reduce dimensionality while retaining important information. This structure enables the system to handle complex patterns in high-dimensional data, such as images, time-series signals, or structured datasets. The method optimizes the neural network's performance by leveraging the hierarchical organization, reducing computational overhead and enhancing feature extraction. The CNN's ability to learn hierarchical representations makes it particularly effective for tasks like image recognition, object detection, and sequence analysis. The overall approach improves the efficiency and accuracy of data processing by combining the strengths of hierarchical algorithms with the deep learning capabilities of CNNs.
11. The method of claim 1 , wherein the lower quality visual data includes a sequence of video frames.
A system and method for processing visual data, particularly for improving the quality of lower-quality visual data such as video frames. The invention addresses the challenge of enhancing visual data that may be degraded due to factors like low resolution, noise, or compression artifacts. The method involves analyzing the lower-quality visual data, which may include a sequence of video frames, to identify and correct distortions or imperfections. Techniques such as frame interpolation, noise reduction, or super-resolution may be applied to improve the visual quality. The system may also incorporate machine learning models trained on high-quality reference data to predict and reconstruct missing or corrupted details in the lower-quality frames. The method ensures that the processed visual data retains temporal coherence, especially when dealing with sequential frames, to avoid unnatural artifacts. The invention is applicable in various fields, including video streaming, surveillance, and medical imaging, where enhancing visual data quality is critical for accurate analysis and user experience.
12. The method of claim 1 , wherein the selecting includes executing a patch-based analysis of the lower quality visual data, the patch-based analysis including comparing features of patches of the fidelity visual data to features of the lower quality visual data.
This invention relates to image processing techniques for enhancing lower quality visual data using higher fidelity reference data. The problem addressed is improving the quality of degraded or low-resolution images by leveraging higher-quality reference images, such as those captured under better conditions or with superior sensors. The method involves selecting corresponding regions between the lower quality visual data and the higher fidelity visual data. This selection process includes a patch-based analysis, where small segments (patches) of the lower quality image are compared to patches in the higher fidelity image. Features within these patches, such as texture, edges, or color patterns, are analyzed to identify matching regions. By aligning and transferring information from the higher fidelity patches to the corresponding lower quality patches, the overall quality of the degraded image is improved. This approach is particularly useful in applications like medical imaging, surveillance, or satellite imagery, where reference data may be available to enhance lower quality captures. The technique ensures that the enhancement is locally accurate, preserving structural and contextual details while mitigating artifacts.
13. The method of claim 1 , further comprising: extracting, by a neural network, frequency features from the fidelity visual data, wherein the transposing the fidelity visual data onto the content of the extracted features of the lower quality visual data includes transposing the frequency features of the fidelity visual data into the content of the extracted features of the lower quality visual data.
This invention relates to enhancing lower quality visual data using frequency features extracted from higher fidelity visual data. The method addresses the challenge of improving the quality of visual data, such as images or video, by leveraging frequency-domain information from a higher-quality reference. A neural network processes the fidelity visual data to extract frequency features, which are then integrated into the content of the lower quality visual data. This transposition of frequency features improves the perceptual quality of the lower quality visual data by incorporating high-frequency details from the fidelity data. The neural network may use techniques such as convolutional layers or attention mechanisms to identify and transfer these features effectively. The method ensures that the enhanced visual data retains the structural and contextual information of the original lower quality data while benefiting from the finer details of the higher fidelity reference. This approach is particularly useful in applications like image super-resolution, video enhancement, or medical imaging, where preserving high-frequency details is critical for accurate interpretation. The technique can be applied in real-time processing systems or offline enhancement pipelines, depending on computational constraints.
14. The method of claim 1 , further comprising: extracting, using a first neural network trained to predict content features of higher quality visual data from the lower quality visual data, content features from the lower quality visual data; extracting, using activations of different layers in a second neural network trained for object recognition, frequency features of the fidelity visual data from the fidelity visual data, wherein the higher quality visual data is estimated using a cost function, wherein the cost function is defined for the higher quality visual data as a sum of: a squared error between content features extracted from a section of higher quality visual data using the activations of different layers in the second neural network and the extracted content features of the lower quality visual data, and a squared error between frequency features extracted from a section of higher resolution visual data using activations of the different layers in the second neural network and the extracted frequency features of the fidelity visual data.
This invention relates to enhancing the quality of visual data, particularly for improving lower-quality visual inputs by leveraging neural networks to estimate higher-quality visual data. The method addresses the challenge of reconstructing high-fidelity visual data from lower-quality inputs, such as compressed or noisy images, by combining content and frequency features extracted from different neural networks. A first neural network is trained to predict content features from lower-quality visual data, capturing essential structural and semantic information. A second neural network, trained for object recognition, extracts frequency features from fidelity visual data, representing high-frequency details and textures. The higher-quality visual data is estimated using a cost function that balances two squared error terms: one between the content features of the lower-quality data and those of the estimated higher-quality data, and another between the frequency features of the fidelity data and those of the estimated higher-quality data. This approach ensures that the reconstructed visual data retains both structural coherence and fine details, improving overall visual quality. The method is particularly useful in applications requiring high-fidelity visual reconstruction, such as medical imaging, surveillance, and multimedia processing.
15. An apparatus comprising: at least one processor; and at least one memory including computer program code which, when executed by the at least one processor, causes the at least one processor to: receive input visual data representing at least one image, the input visual data including lower quality visual data; select fidelity visual data from a database of fidelity visual data, the fidelity visual data including a plurality of features, the plurality of features being computed by a neural network trained on object recognition; extract features from the lower quality visual data; and estimate, by a hierarchical algorithm, higher quality visual data from the lower quality visual data and the fidelity visual data, including transpose the plurality of features of the fidelity visual data onto content of the extracted features of the lower quality visual data, the higher quality visual data having a resolution higher than the lower quality visual data.
This invention relates to image enhancement using neural networks and hierarchical algorithms. The problem addressed is improving the quality of low-resolution or degraded visual data by leveraging high-fidelity reference images. The apparatus includes at least one processor and memory storing computer program code. The system receives input visual data, such as an image, which is of lower quality. It then selects fidelity visual data from a database, where this data includes multiple features computed by a neural network trained for object recognition. Features are extracted from the lower-quality input data. A hierarchical algorithm then estimates higher-quality visual data by combining the lower-quality input with the fidelity data. Specifically, the algorithm transposes the features from the fidelity data onto the content of the extracted features from the lower-quality data, resulting in an output with higher resolution than the original input. The neural network ensures that the features used for enhancement are relevant to object recognition, improving the accuracy and coherence of the enhanced image. This approach is particularly useful in applications requiring real-time image enhancement, such as video streaming, medical imaging, or surveillance systems.
16. The apparatus of claim 15 , wherein the database includes a plurality of feature vectors, wherein the at least one memory includes computer program code that when executed by the at least one processor causes the at least one processor to: compute a feature vector for the lower quality visual data, the feature vector configured to approximate a feature vector of the higher quality visual data; and compare the feature vector to the plurality of feature vectors of the database to obtain visual data similar to the lower quality visual data by maximizing a cosine similarity between the feature vector and one or more of the plurality of feature vectors included in the database, the visual data including the plurality of features of the fidelity visual data.
This invention relates to visual data processing, specifically improving the quality of lower-quality visual data by comparing it to higher-quality reference data stored in a database. The problem addressed is the difficulty in enhancing low-resolution or degraded visual data without access to high-fidelity reference material. The solution involves generating a feature vector from the lower-quality visual data that approximates the feature vector of higher-quality visual data. This feature vector is then compared to a database of precomputed feature vectors derived from high-fidelity visual data. The comparison is performed by maximizing cosine similarity between the computed feature vector and the stored feature vectors, identifying the most similar high-fidelity visual data. The system includes a database containing multiple feature vectors, a processor, and memory storing executable code. The processor computes the feature vector for the input visual data and performs the similarity comparison to retrieve the closest matching high-fidelity visual data. This approach enables the reconstruction or enhancement of lower-quality visual data by leveraging pre-existing high-quality references, improving visual fidelity without requiring direct access to the original high-quality source.
17. The apparatus of claim 15 , wherein the plurality of features of the fidelity visual data include one or more content features and one or more style features, the features of the lower quality visual data are extracted using the hierarchical algorithm.
This invention relates to visual data processing, specifically improving the fidelity of lower-quality visual data by extracting and applying features from higher-quality visual data. The problem addressed is the degradation of visual quality in lower-resolution or compressed images, which often results in loss of detail, color accuracy, and stylistic elements. The solution involves an apparatus that processes visual data by analyzing both content and style features hierarchically to enhance the lower-quality input. The apparatus extracts features from the lower-quality visual data using a hierarchical algorithm, which systematically decomposes the data into multiple levels of abstraction. Content features represent the structural or semantic elements of the visual data, such as edges, textures, and object shapes. Style features capture aesthetic attributes like color palettes, lighting, and artistic effects. By separating these features, the apparatus can selectively enhance or modify them to improve fidelity. The hierarchical approach ensures that both high-level and low-level features are considered, allowing for precise adjustments that maintain coherence between different parts of the visual data. This method is particularly useful in applications like image upscaling, video enhancement, and style transfer, where preserving both content integrity and stylistic consistency is critical. The apparatus may be implemented in software, hardware, or a combination thereof, depending on the specific use case.
18. A non-transitory computer-readable medium having computer-readable code stored thereon, the computer-readable code, when executed by at least one processor, causes the at least one processor to: receive input visual data representing at least one image, the input visual data including lower quality visual data; select fidelity visual data from a database of fidelity visual data based on similarity with the lower quality visual data, wherein the similarity is based on maximizing a cosine similarity function; extract features from the lower quality visual data; and estimate, by a hierarchical algorithm, higher quality visual data from the lower quality visual data and the fidelity visual data, including execute a fidelity transfer by transposing the fidelity visual data onto content of the extracted features of the lower quality visual data, the higher quality visual data having a resolution higher than the lower quality visual data.
This invention relates to image enhancement techniques for improving the quality of low-resolution or low-fidelity visual data. The problem addressed is the degradation of visual data due to factors such as compression, noise, or limited sensor capabilities, which results in lower resolution or poor visual quality. The solution involves a computational method that enhances the fidelity of input visual data by leveraging a database of high-quality reference images. The system receives input visual data, which may be of lower quality, and compares it to a database of high-fidelity visual data. The comparison is performed using a cosine similarity function to identify the most similar high-fidelity reference. Features are then extracted from the lower-quality input data. A hierarchical algorithm processes these features along with the selected high-fidelity reference to generate enhanced visual data. The enhancement includes a fidelity transfer step, where the high-quality characteristics of the reference data are transposed onto the content of the input data. The output is higher-resolution visual data, improving clarity and detail compared to the original input. This approach is particularly useful in applications requiring high-quality visual output from limited or degraded input sources.
19. The non-transitory computer-readable medium of claim 18 , wherein the database includes a plurality of feature vectors, wherein the computer-readable code further includes computer-readable code that, when executed by the at least one processor, causes the at least one processor to: compute a feature vector for the lower quality visual data, the feature vector configured to approximate a feature vector of the higher quality visual data; and compare the feature vector to the plurality of feature vectors of the database to obtain visual data similar to the lower quality visual data.
This invention relates to improving visual data quality by leveraging a database of feature vectors to enhance lower-quality visual data. The system addresses the challenge of processing low-resolution or degraded visual inputs by approximating their high-quality counterparts using machine learning techniques. The database contains multiple feature vectors derived from high-quality visual data, which serve as reference points for comparison. The system computes a feature vector for the input lower-quality visual data, designed to closely match the structure of feature vectors from higher-quality data. By comparing this computed feature vector against the database, the system identifies visually similar high-quality data, enabling enhancements such as resolution improvement or noise reduction. The approach relies on feature vector similarity to bridge the gap between low and high-quality visual representations, improving the usability of degraded visual inputs in applications like image processing, computer vision, and multimedia analysis. The method ensures that the feature vector computation and comparison are optimized for accuracy and efficiency, leveraging pre-existing high-quality data to enhance lower-quality inputs.
20. The non-transitory computer-readable medium of claim 18 , wherein the database includes a plurality of feature vectors, wherein the computer-readable code further includes computer-readable code that, when executed by the at least one processor, causes the at least one processor to: extract, using a first neural network trained to predict content features of higher quality visual data from the lower quality visual data, content features from the lower quality visual data; and extract, using activations of different layers in a second neural network trained for object recognition, frequency features of the fidelity visual data from the fidelity visual data, wherein the higher quality visual data is estimated using a cost function, wherein the cost function is defined for the higher quality visual data as a sum of: a squared error between content features extracted from a section of higher quality visual data using the activations of different layers in the second neural network and the extracted content features of the lower quality visual data, and a squared error between frequency features extracted from a section of higher resolution visual data using activations of the different layers in the second neural network and the extracted frequency features of the fidelity visual data.
This invention relates to enhancing the quality of visual data using neural networks. The problem addressed is improving the fidelity of lower quality visual data by leveraging higher quality reference data and neural network-based feature extraction. The system includes a database containing feature vectors and employs two neural networks. The first neural network is trained to predict content features of higher quality visual data from lower quality input. The second neural network, trained for object recognition, extracts frequency features from fidelity visual data by analyzing activations across different layers. The system estimates higher quality visual data using a cost function that combines two squared error terms. The first term measures the difference between content features extracted from higher quality data and those from the lower quality input, both processed by the second neural network. The second term compares frequency features extracted from higher resolution data and those from the fidelity data, again using the second neural network. This approach ensures that the enhanced visual data retains both content and frequency characteristics of higher quality references.
Unknown
November 3, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.